GH-38914: [Python] Add EncryptionConfiguration.uniform_encryption

[MartinNowak]

Rationale for this change

Support uniform encryption when writing parquet files.

What changes are included in this PR?

Exposing EncryptionConfiguration.uniform_encryption to pyarrow.

Are these changes tested?

Yes, see included tests.

Are there any user-facing changes?

Addition of uniform_encryption parameter to EncryptionConfiguration.

• fixes [Python][Parquet] Add EncryptionConfiguration.uniform_encryption to Python implementation #38914
• supersedes GH-38914: [Python] Add EncryptionConfiguration.uniform_encryption #39216
• GitHub Issue: [Python][Parquet] Add EncryptionConfiguration.uniform_encryption to Python implementation #38914

[MartinNowak]

Python / AMD64 Conda Python 3.10 Without Pandas (pull_request) Failing after 17m

GDB test failures seem unrelated.

[rok]

Thanks for picking this up @MartinNowak. PR looks good to me.
Only one nit - if I understand correctly uniform_encryption defaults to False so encrypting with:

    encryption_config = pe.EncryptionConfiguration(
        footer_key=FOOTER_KEY_NAME,
        encryption_algorithm="AES_GCM_V1",
        cache_lifetime=timedelta(minutes=5.0),
        data_key_length_bits=256)

Should raise: "Either column_keys or uniform_encryption must be set". Could we add test for that?

[MartinNowak]

Should raise: "Either column_keys or uniform_encryption must be set". Could we add test for that?

Thanks for the quick review @rok. That was already tested.

arrow/python/pyarrow/tests/parquet/test_encryption.py

Lines 253 to 267 in b78e6e2

	def test_encrypted_parquet_write_no_col_key(tempdir, data_table):
	"""Write an encrypted parquet, but give only footer key,
	without column key."""
	path = tempdir / 'encrypted_table_no_col_key.in_mem.parquet'
	# Encrypt the footer with the footer key
	encryption_config = pe.EncryptionConfiguration(
	footer_key=FOOTER_KEY_NAME)
	with pytest.raises(OSError,
	match="Either column_keys or uniform_encryption "
	"must be set"):
	# Write with encryption properties
	write_encrypted_file(path, data_table, FOOTER_KEY_NAME, COL_KEY_NAME,
	FOOTER_KEY, b"", encryption_config)

.

[rok]

Thanks for the quick review @rok. That was already tested.

Oh yeah, you're right.

[rok]

This looks good to go.
Pinging @pitrou to double check me :).

[rok]

The gdb issue seems unrelated.

[rok]

Assuming CI passes I'll merge @pitrou

[pitrou]

LGTM, thanks a lot for this @MartinNowak

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 6ab37dd.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 3 possible false positives for unstable benchmarks that are known to sometimes produce them.